Overlapping MTRRs in 2.6.1

From: Ryan Reich
Date: Sun Jan 18 2004 - 16:14:23 EST


I get the following message in the kernel log when X starts up:

mtrr: 0xe0000000,0x4000000 overlaps existing 0xe0000000,0x1000000
mtrr: 0xe0000000,0x4000000 overlaps existing 0xe0000000,0x1000000
[drm:radeon_cp_init] *ERROR* radeon_cp_init called without lock held
[drm:radeon_unlock] *ERROR* Process 1807 using kernel context 0

The complaint with radeon_unlock occurred in 2.4 for a while until I realized
that agpgart wasn't loading before radeon; that fixed it, but now it's returned
with 2.6 and I think the mtrr error is the cause. In XFree86.log, I get the
single-line message

(WW) RADEON(0): Failed to set up write-combining range (0xe0000000,0x4000000)

and then, further down,

(WW) RADEON(0): [agp] AGP not available

and finally,

(II) RADEON(0): Direct rendering disabled

I checked the archives and no one has posted a message on MTRR overlaps since
2002. At that time, Andrew Rodland wrote something about what appears to be
this exact problem; a link to it is

http://marc.theaimsgroup.com/?l=linux-kernel&m=102133293210888&w=2

It would seem this hasn't gotten fixed entirely, then. For comparison, 'cat
/proc/mtrr' gives

reg00: base=0x00000000 ( 0MB), size= 256MB: write-back, count=1
reg01: base=0xe8000000 (3712MB), size= 64MB: write-combining, count=1
reg02: base=0xe0000000 (3584MB), size= 16MB: write-combining, count=1

My video card is a Radeon 7000, 64M of memory; processor, Athlon 2600+;
motherboard, Shuttle AN35N with NForce2 chipset.

--
Ryan Reich
ryanr@xxxxxxxxxxxx
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/